AdaExploreBench (Forced Probe) - Calibrated Accuracy: leaderboard

AdaExploreBench's calibrated accuracy when probing is mandatory in every exploration window with a minimum number of probe cases, a diagnostic setting that separates deciding to explore from using what exploration reveals.

Metric: CA (%) (0-100). Source: hkust-knowcomp.github.io. Saturation forecast: Around March 2027. 6 models tracked. Updated 19 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-6 SolHigh53.99#33
2DeepSeek V4.1 FlashHigh53.27#69
3Qwen 3.8 27BMedium38.69#220
4GPT-6 LunaHigh32.27#106
5Gemma 4 12B (IT)High22.02#731
6Qwen 3.5 9B14.32#829

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=adaexplorebench-forced-probe-calibrated-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-20.